JP3207281B2 - Stereo speech encoding / decoding system, stereo speech decoding device, and single speech / multiple simultaneous speech discrimination device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention, stereo audio encoding and decrypt scheme applied to the teleconference system or the like, stereo
The present invention relates to a voice decoding device and a single message / multiple simultaneous message discriminating device .

[0002]

2. Description of the Related Art In recent years, with the advance of communication technology, a demand for a remote conference system for holding a conference by communication between remote locations has been increasing.

As shown in FIG. 5, such a remote conference system generally includes an input / output system such as a microphone 1, a speaker 2, a TV camera 3, a TV 4, an electronic whiteboard 5, a FAX 6, and a telewriting 7, and an audio unit 8. , Control unit 9,
It is composed of a control system such as a control pad 10 and an image unit 11, and a transmission system such as a transmission path 12 and a transmission unit 13.
Image information and audio information such as moving images and still images are exchanged between remote locations via the transmission path 12.

By the way, in such a teleconference system, it is desired to reduce the transmission cost. In particular, if such information can be transmitted at a transmission speed of about 64 kbps which can be transmitted through a current general subscriber line, an optical fiber or the like is used. A teleconference system can be realized at a much lower cost than a high-quality teleconference system that has been used.

In a teleconferencing system using such a transmission line with a low transmission rate, even a monaural voice can be compressed to 16 kb by voice data compression processing such as ADPC.
Since it is necessary to compress the data to a low transmission capacity of about ps, stereo sound is not usually used.

[0006] However, it is well known that in a teleconferencing system, it is desirable to employ stereophonic sound in order to give a sense of realism and to identify a speaker who knows who is speaking.

Accordingly, the present inventors have disclosed in Japanese Patent Application Laid-Open No. Sho 62-51.
Japanese Patent Publication No. 844 proposes a stereo audio transmission system that enables high-quality stereo audio transmission at low cost even when used in a transmission line with a low transmission rate.

This method will be briefly described with reference to FIG.

[0009] As shown in FIG.
(Ω) is input to the microphones 1r and 1l of the left and right channels. However, an echo from a wall or the like is ignored, and ω represents each frequency. At this time, if the transfer functions of the left and right channels are G _L (ω) and G _R (ω), the input voices Y _L (ω) and Y _R (ω) of the left and right channels are Y _L (ω) = the _{G L (ω) X (ω} ) ...... (1) Y R (ω) = G R (ω) X (ω) ...... (2). Moreover, from both equations _{Y L (ω) = (G} L (ω) / G R (ω)) · Y R (ω) ...... (3) = G (ω) Y R (ω) ...... (4) Becomes

Therefore, if the transfer function G (ω) is known, the sound of the right channel can be reproduced.

Therefore, in this system, in the case of stereo transmission, the transmission function G (ω) estimated as the audio signal Y _R (ω) of one channel from the transmission side without transmitting the audio of both channels independently. The transmitting and receiving sides combine the audio signal Y _R (ω), the audio signal Y _R (ω), and the transfer function G (ω) to convert the audio of both channels into the left and right speakers 2.
By performing reproduction from r and 2l, stereo transmission is performed.

[0012] In this method, since a transfer function can be defined simply by delay and attenuation if a single utterance is assumed, the information amount is much smaller than the information amount of the voice signal Y _L (ω). Estimation can be easily performed, and stereo transmission can be performed with a smaller transmission amount.

However, this method is based on the premise that a single utterance is made. Therefore, an accurate transfer function G (ω), that is, additional information, can be generated during a double talk in which a plurality of speakers speak simultaneously. There was a problem that the sound image fluctuated.

[0014]

In a conversation such as a conference, it is generally considered that the ratio of double talk is very low. In the conventional method, large bandwidth compression was realized by transmitting a single message monaurally using this property. is there.

[0015] Therefore, the present invention provides high-quality stereo audio encoding and decrypt schemes were not unsteady of sound even during double talk, stereo audio decoding apparatus and alone statements / multiple simultaneous
An object of the present invention is to provide a speech discrimination device .

[0016]

To solve SUMMARY OF THE INVENTION The above problem, stereo audio encoding and decrypt method of the first invention,
Stereo for encoding and decrypt the audio signals of a plurality of channels
The voice encoding / decoding method has a function of distinguishing a single utterance or a plurality of simultaneous utterances. At the time of a single utterance, the main information including the audio signal of at least one of the plurality of audio signals and the main information turned into encoding and decrypt the additional information required to synthesize the speech signal of the remaining channels from the information, when multiple simultaneous utterance, characterized in that the encoding and decrypt individual audio signals of the plurality of channels And

The stereo audio coding and decrypt method of the second aspect, in the first aspect, the encoded voice band of the main information encoding method when alone remarks, encoding each of the at multiple simultaneous speaking It is wider than the voice band.

A third stereo audio decrypt apparatus of the aspect of the present invention,
Means for decrypt and additional information necessary for synthesizing the audio signal of the remaining channels from the main information with the main information consisting of audio signals of the at least one channel of the plurality of channels of audio signals, said plurality of channels means for decrypt separately an audio signal, based on the additional information,
And means for distinguishing single utterance or more concurrent speaking, when alone speaking, select the primary information and the additional information which is the decrypt, the multiple simultaneous utterance, means for selecting said individually Fukugo of information It is characterized by having.

In a fourth aspect of the present invention, a single utterance / multiple simultaneous utterance discriminating apparatus inputs an audio signal to first and second low-speed filters.
Means from the first and second low-speed filters.
Left and right sums obtained by adding and subtracting the respective output signals
Adaptive prediction (ADPCM) code for each component and difference component
Means for generating a sign component turned into, is the product
Means for inputting the sign component delay line, or the delay line
Averaging is performed on these output signals to obtain the code phase between T samples.
Correlation generating means for obtaining the correlation, and T correlation by the correlation generating means.
It is characterized by comprising a judgment means for judging a plurality of simultaneous utterances when a correlation output cannot be obtained between samples, and for judging a single utterance when obtaining the correlation output .

[0020]

[0021]

[0022]

According to the present invention, stereo sound transmission is performed at the time of double talk, and only a single utterance is transmitted in monaural, thereby preventing sound image fluctuation. However, if stereo transmission is simply performed during double talk, the transmission rate will increase even temporarily during double talk. Therefore, by slightly deteriorating the quality only during double talk,
Stereo transmission is realized without increasing the transmission rate.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a diagram showing a configuration of a system according to an embodiment of the present invention.

As shown in FIG.
It comprises a third monaural speech encoder 101, 103, 104 and a detector 102.

The first monaural speech coder 101 compresses the sum of the outputs of the left and right microphones to, for example, 56 kbps when coding alone, and codes the sum. The detector 102 detects a level difference between the left and right microphone output signals, a delay difference, and a difference between a single utterance or a plurality of simultaneous utterances, and encodes them at, for example, 8 kbps. Second and third monaural audio encoders 103, 1
Reference numeral 04 individually encodes the left and right microphone output signals at a lower bit rate, for example, 32 kbps, by the first monaural audio encoder 101 at the time of simultaneous speech.

The decrypt unit side, the first to third mono decrypt <br/> encoder 105,108,109, pseudo-stereo generator 10
6 and 107 and selectors 110 and 111.

The first mono decrypt 105 decodes the code sent from the first monaural audio encoder 101 when alone speak. Pseudo stereo generators 106, 10
7 pseudo-generates stereo sound by giving a delay difference and a gain difference to the decoded output. Second and third monaural reproduction
The encoders 108 and 109 decode the left and right codes sent from the second and third monaural audio encoders 103 and 104 at the time of simultaneous speech. Selectors 110, 111
Alone speak, on the basis of the determination result of multiple simultaneous utterance, and selects and outputs either the output of the output or the second and third mono decrypt 108 and 109 of the pseudo-stereo generators 106 and 107 .

According to this configuration, at the time of a single utterance that occupies most of the conversation, the first monaural speech encoder 101 can perform pseudo-quality stereo speech transmission at a transmission rate of, for example, 64 kbps with high quality. Also, at the time of plural simultaneous speeches and other states, the second and third monaural audio encoders 103 and 104 separately set the left and right
Complete stereo audio transmission encoded at 2 kbps can be performed. As a result, although the quality is slightly degraded compared to a single utterance, it is possible to perform coded transmission at a total of 64 kbps, to prevent disturbance of sound images when a plurality of utterances are simultaneously uttered while keeping the total coding speed constant, and to achieve high quality communication when a single utterance is made. it can.

Next, the configuration of each section will be described in detail. In the following, it is assumed that the 7 kHz band wideband speech coding system is applied to a single speech, and the 3.4 kHz band telephone band speech coding system is applied to a plurality of simultaneous speeches and other states.

FIG. 2 shows an example of the configuration on the encoder side.

As shown in the figure, the left and right microphone output sounds are divided into band division filters 301, 302, 303, 304.
0 to 4 kHz (0 to 4 kHz at the same time)
The frequency band is divided into a low band of 3.4 kHz) and a high band of 4 to 7 kHz.

Of these filter outputs, the outputs of the high-pass filters 301 and 304 are added to the left and right signals by an adder 305, and then coded to 16 kbps by an adaptive prediction (ADPCM) coder 306. It becomes a part of the transmission data at the time of a single remark. The low-pass filter 30
2 and 303 are output from an adder 307 and a subtractor 308.
Are combined as left and right sum components and difference components, and input to the ADPCM encoders 309 and 310, respectively. Of these, the sum component is converted to 4 by the ADPCM encoder 309.
The data is encoded at 0 kbps and becomes a part of the transmission data at the time of a single utterance.
One bit of the SB is removed, and together with the output of the difference component ADPCM encoder 310, becomes the transmission data at the time of plural simultaneous utterances of 32 kbps.

Further, ADPCM encoders 309 and 31
The sign component of the output of 0 and the input signal are calculated by the estimator 31.
At the same time, a left-right level difference and a delay difference are detected, and at the same time, a single utterance and a plurality of simultaneous utterances are determined.

The single comment data synthesizer 312 is 16 kb.
A 64 kbps transmission data is generated by combining a ps ADPCM high band code, a 40 kbps low band sum component ADPCM code, and an 8 kbps estimator 311 output code.

A plurality of simultaneous speech synthesizers 313 are 32 kbp.
The output codes of the left and right ADPCM encoders 306 and 310 are combined to generate transmission data of 64 kbps.

The transmission data is selected by a switch 314 from a transmission data sequence of 64 kbps by selecting a single transmission data sequence based on a single plural-speech determination signal output from the estimator 311.
Is sent to the line.

[0038] FIG. 3 is a configuration example of a decrypt unit side.

As shown in the figure, a received data sequence of 64 kbps is input to a distributor 315 for single speech and a distributor 316 for simultaneous speech.

4 out of the outputs of the distributor 315 for a single statement
The 0 kbps ADPCM code is input to an ADPCM decoder 317 for low frequency, and the 16 kbps ADPCM code is input to an ADPCM decoder 318 for high frequency. The output of these decoders is the pseudo-stereo synthesizer 31
9, 320, 321, and 322, are generated into left and right pseudo stereo sound based on the 8 kbps output of the distributor 315, which is the delay difference gain difference detected on the encoder side.
Low-pass filters 323 and 324 of band 0.2 to 4 kHz (3.4 kHz at the time of simultaneous speech) for band synthesis, band 4
It is input to high-pass filters 325 and 326 of up to 7 kHz. Outputs of these filters are band-synthesized by adders 327 and 328, and then become decoded signals at the time of a single utterance.

On the other hand, two 32-kbps data sequences output from the distributor 316 for simultaneous simultaneous speech are decoded by the low-frequency ADPCM decoders 317 and 326, and then left and right from the sum component difference component. Adder 33 that restores the signal of
0 and input to the subtractor 331. These outputs are
The signals are input to the low-pass filters 323 and 324 for band synthesis only when a plurality of messages are sent simultaneously by the switches 332 and 333.

ADPCM decoders 317, 32 for low band
The positive / negative sign component of the input code of No. 6 is input to the detector 334 and is used as a signal for switching from the multiple simultaneous speech state to the single speech state.

The switches 335 and 336 are used to suppress high-frequency components that cannot be decoded at the time of simultaneous speech.

FIG. 4 shows an example of the configuration of the estimator 311.

As shown in FIG.
One signal SIGN (R) (the right component in this example) of the positive and negative sign components of the encoders 309 and 310 is input to the delay line 401 with taps for N samples. On the other hand, the other sign (the left component in this example) is input to the N / 2 sampled delay line 402 for establishing the right and left causality.
Output signals of these delay lines are output to exclusive OR circuits 403-1,. . 403-
, N, and then cleared every T samples. . . . The code correlation between T samples is obtained by performing the averaging process by 404-N.

The up / down counter 404-
1,. . . . The output of 404-N is latched by the latch 405 immediately before clearing, and then encoded by the decoder circuit 406 to become left and right delay difference information τ updated every T samples.

The timer 407 generates a clear signal CL and a latch signal LTC for every T samples. Generally, T is set to a value of, for example, about 100 msec.

A code corresponding to all the outputs of the latch circuit 405 out of the outputs of the decoder circuit 406 corresponds to the OR circuit 4
At 08, a state in which no correlation output is obtained between T samples, that is, a plurality of simultaneous speech states is determined.

[0049] The above circuit, the decrypt-side detector 334
A switching signal to a single utterance of a plurality remarks in decrypt instrument used to.

On the encoder side, the level detector 4
09, 410 and the comparator 411 detect the left / right level ratio 1 and become additional information together with the delay difference.

[0051] Thus, in this embodiment, stereo audio encoding method that does not disturbance of the sound image when multiple simultaneous speech by adding a relatively simple process to wideband mono ADPCM encoder-decrypt unit which is widely used realizable.

The above embodiment is merely an example for carrying out the present invention, and various modifications can be made without departing from the spirit of the present invention.

[0053]

As described above, according to the present invention,
Since stereo sound transmission is performed during double talk and monaural sound transmission is performed only for a single utterance, fluctuation of a sound image can be prevented, and high quality stereo sound can be realized.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of a system according to an embodiment of the present invention.

2 is a configuration example of a mark-encoder side shown in FIG.

3 is a configuration example of a decrypt unit side shown in FIG.

FIG. 4 is a configuration example of an estimator shown in FIG . 2 ;

FIG. 5 is a diagram showing a general configuration of a remote conference system.

FIG. 6 is a diagram for explaining a conventional stereo sound transmission system.

[Explanation of symbols]

101, 103, 104 ... first to third mono audio coder 102 ... detector 105,108,109 ... first to third mono decrypt 106 and 107 ... pseudo-stereo generator 110, 111 ... Selector

Continuation of the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) H04B 14/04 H04N 7/15 H04M 3/56

Claims (4)

(57) [Claims] An audio signal of a plurality of channels is encoded and decoded.
In the stereo audio encoding / decoding method to be encoded, the audio signal has a function of distinguishing a single utterance or a plurality of simultaneous utterances. When the single utterance is made, the main information including the audio signal of at least one channel among the audio signals of the plurality of channels. turned into encoding and decrypt and additional information necessary for synthesizing the remaining channels of the audio signal from the main information of Toko, when multiple simultaneous utterance, to encoding and decrypt individual audio signals of the plurality of channels Stereo speech coding
Decrypt method. 2. A stereo speech encoding / decoding method according to claim 1,
Alone encoding voice band of the main information encoding method when remarks, stereo audio encoding and decrypt scheme, wherein the wider encoded voice band of each time multiple simultaneous speech at No. scheme. Wherein means for decrypt and additional information necessary for synthesizing the audio signal of the remaining channels from the main information with the main information consisting of audio signals of the at least one channel of the plurality of channels of audio signals If, means for decrypt individual audio signals of the plurality of channels, based on the additional information, and means for distinguishing single utterance or more simultaneous utterance, alone at the time of speaking, the decrypt been main information added selects information, the multiple simultaneous speaking, the individual stereo audio decrypt apparatus characterized by comprising a means for selecting Fukugo of information. 4. The method according to claim 1, wherein the audio signal is supplied to first and second low-speed filters.
Input means and respective outputs from the first and second low speed filters
Left and right sum components and difference components obtained by adding and subtracting signals
Minute sign with adaptive prediction (ADPCM) encoding
Means for generating a sign component, and means for inputting the generated sign component to a delay line
When, T sump subjected to averaging processing on an output signal from the delay line
Generating a correlation output between T samples by the correlation generating means for obtaining the code correlation between the T samples.
If not, it is determined that there is a plurality of simultaneous remarks, and the correlation output is
A single utterance / multiple simultaneous utterance discriminating device, comprising: a deciding means for judging a single utterance when obtained .